Luma Labs released Dream Machine, an AI video generation tool. Soon after the announcement, many users came to try out the new image-to-video model. Initially, there was a lot of traffic and all requests were queued, so no videos were generated. Luma Labs made the video model available to all users for free and without waiting. Free users can use it for up to 30 generations per month, after which paid plans are available.
Introducing Dream Machine – a next-gen video model that uses AI to create high-quality, realistic shots from text instructions and images. Available to everyone today. Try it for free here https://t.co/rBVWU50kTc #LumaDreamMachine pic.twitter.com/Ypmacd8E9z
— Luma AI (@LumaLabsAI) June 12, 2024
The quality of their model aims to compete with next-gen video models like Sora and the recently released Chinese model (which currently requires a waiting list). Additionally, Pika Labs has also released an improved model for AI-generated videos. The tool is highly advanced and attempts not only to generate motion videos that animate photos, but also to create entire video scenes. While not always successful, creating videos from uploaded photos works much better than generating them from text prompts.
I follow all the AI news and try to test all the newly released models 🤖
I feel like I have too much access to it already.
* GPT-4o + LumaLab pic.twitter.com/sjoib45Aak
— TestingCatalog News 🗞 (@testingcatalog) June 13, 2024
The UI is very simple, with two columns for a text prompt, a section for uploading an image, and a section at the bottom for viewing your creation. Once you enter the prompt, the video generation is queued and you'll see an estimated remaining time. After the initial load, the team made some modifications, which significantly reduced the time it takes to generate a video; it's usually around 120 seconds, but sometimes it can take a bit longer. The minimal UI is enough to get the job done.
We found that generating video from JPEGs works better than using other formats or text, but it can take several tries to generate something meaningful. The video previewed in the Sora demo was excellent, but there are already many impressive examples from Luma Labs that demonstrate a high level of quality. Let's take a look at performance:
